Influence of Single Impurity Atoms on the Structure, Electronic, and

Feb 28, 2007 - The square−pyramidal isomer of bare Ni5 with six unpaired electrons was calculated 23 kJ/mol more stable than the trigonal bipyramid ...
0 downloads 0 Views 262KB Size
Download PDF
J. Phys. Chem. A 2007, 111, 2067-2076

2067

Influence of Single Impurity Atoms on the Structure, Electronic, and Magnetic Properties of Ni5 Clusters Petko St. Petkov,† Georgi N. Vayssilov,*,† Sven Kru1 ger,‡ and Notker Ro1 sch*,‡ Faculty of Chemistry, UniVersity of Sofia, 1126 Sofia, Bulgaria, and Department Chemie, Theoretische Chemie, Technische UniVersita¨t Mu¨nchen, 85747 Garching, Germany ReceiVed: NoVember 14, 2006; In Final Form: January 16, 2007

With a gradient-corrected density functional method, we have studied computationally the influence of single impurity atoms on the structure, electronic, and magnetic properties of Ni5 clusters. The square-pyramidal isomer of bare Ni5 with six unpaired electrons was calculated 23 kJ/mol more stable than the trigonal bipyramid in its lowest-energy electronic configuration with four unpaired electrons. In a previous study on the cluster Ni4, we had obtained only one stable isomer with an O or an H impurity, but we located six minima for ONi5 and five minima for HNi5. In the most stable structures of HNi5, the H atom bridges a Ni-Ni edge at the base or the side of the square pyramid, similarly to the coordination of an H atom at the tetrahedral cluster Ni4. The most stable ONi5 isomers exhibit a trigonal bipyramidal structure of the Ni5 moiety, with the impurity coordinated at a facet, (µ3-O)Ni5, or at an apex edge, (µ-O)Ni5. We located four stable structures for a C impurity at a Ni5 cluster. As for CNi4, the most stable structure of the corresponding Ni5 complex comprises a four-coordinated C atom, (µ4-C)Ni5, and can be considered as insertion of the impurity into a Ni-Ni bond of the bare cluster. All structures with C and five with O impurity have four unpaired electrons, while the number of unpaired electrons in the clusters HNi5 varies between 3 and 7. As a rough trend, the ionization potentials and electron affinities of the clusters with impurity atoms decrease with the coordination number of the impurity. However, the position of the impurity and the shape of the metal moiety also affect the results. Coordination of an impurity atom leads to a partial oxidation of the metal atoms.

1. Introduction Impurities of nonmetal atoms on small metal clusters, either in the gas phase or on support, can considerably affect the structure, electronic, and magnetic properties of the metal moiety. Such “ligated” clusters can be produced by the interaction of bare metal clusters with gas molecules,1 by reactions of supported metal clusters with surface active sites, e.g., OH groups,2,3 or by the decomposition of molecular ligands (e.g., CO) adsorbed on the metal moiety.4-6 The latter approach is routinely applied in the preparation of supported metal species on oxide surfaces or in zeolite cages when carbonylated transition metal clusters are used as precursors.6-8 In several computational studies, complementing available experimental information, we have concluded that supported metal clusters produced in this way are not free of heteroatoms.2,3,5 Despite the importance of such impurity atoms for the behavior of metal clusters, experimental information about the presence of heteroatoms and their influence on the properties of clusters is rather scarce due to difficulties with preparing and characterizing such samples in a reproducible way. This lack of essential experimental information provides one of the motivations for the theoretical model studies of small transition metal clusters. Various computational approaches have been applied to clarify bonding in and magnetic properties of bare metal clusters;9-24 alternatively, one characterizes clusters with impurity atoms via spectral features.2,3,5,25-31 * Corresponding authors. E-mails: [email protected]; roesch@ ch.tum.de. † University of Sofia. ‡ Technische Universita ¨ t Mu¨nchen.

Recently, we reported a computational study of Ni4 clusters containing single impurity atoms H, C, or O.32 These impurity atoms were selected not only according to their importance for the gas-phase chemistry of size-selected clusters, but also with regard to heterogeneous catalytic systems where supported transition metal clusters often are utilized as components of active multifunctional catalysts.33,34 We determined unique stable isomers of the species HNi4 and ONi4 where the impurity atom adopts a 2-fold coordinated position at an edge of the tetrahedral Ni4 cluster. We also located three structures for the cluster CNi4 where C exhibits a preference for 4-fold coordination. The bonding of the impurity atoms H, C, or O to the cluster causes a partial oxidation of the metal atoms. In the present work, we considered how single impurity atoms, H, C, or O, affect the structure and other properties of the cluster Ni5. As mentioned above, these impurities often are present in heterogeneous catalytic systems, either from the preparation process of the catalyst or from catalytic reactions. We determined the magnetic states of such clusters, their spin and electron density distribution as well as their ionization potential (IP) and electron affinity (EA). 2. Computational Details All calculations were carried out with the linear combination of Gaussian-type orbitals fitting-functions density functional method (LCGTO-FF-DF)35 as implemented in the program PARAGAUSS.36,37 We applied the unrestricted Kohn-Sham procedure with the gradient-corrected exchange-correlation functional suggested by Becke (exchange) and Perdew (correlation).38,39 A recent extensive study on transition metal

10.1021/jp0675431 CCC: $37.00 © 2007 American Chemical Society Published on Web 02/28/2007

2068 J. Phys. Chem. A, Vol. 111, No. 11, 2007

Petkov et al.

TABLE 1: Calculated Characteristics of Bare Ni5 and Ni5 Clusters with H, C, and O Impurity Atoms:a Number Ns of Unpaired Electrons, Binding Energy BE per Atom of Bare Ni5, BE of Impurity Atoms at Ni5 (Both in kJ/mol), Interatomic Distances (in pm)

c

structures

Ns

BE

Ni-Nib

Ni-Xc

Ni5/C4V-sp Ni5/C3V-bp (µ-H)Ni5/Cs-sp1 (µ-H)Ni5/Cs-sp2 (µ3-H)Ni5/Cs-bp (η-H)Ni5/Cs-bp (η-H)Ni5/Cs-sp (µ4-C)Ni5/Cs-sp (µ3-C)Ni5/Cs-sp (η-C)Ni5/Cs-bp (η-C)Ni5/Cs-sp (µ3-O)Ni5/Cs-bp (µ-O)Ni5/Cs-bp (µ-O)Ni5/C1-sp (µ3-O)Ni5/Cs-sp (µ4-O)Ni5/Cs-sp (η-O)Ni5/Cs-sp

6 4 5 5 5 3 7 4 4 4 4 4 4 6 4 4 4

231 226 273 266 262 234 209 660 646 403 378 436 412 402 390 340 313

231 aa ab 237 ab, 229 bb, 236 bc 231 ab ac, 234 bc, 225 bb, 234 cc 242 ab, 232 ac, 231 ad, 235 bd, 233 cd 232 ab, 243 ad, 233 bc, 236 bd, 234 dc, 225 dd 239 ab, 234 ad, 231 bc, 231 bd, 235 cd, 230 dd 234 ab, 235 ac, 231 cd 232 ab, 253 ad, 231 bd, 241 cd 237 ab, 235 ac, 237 ad, 231 bd, 243 cd 234 ab, 228 ac, 233 ad, 227 bd, 246 cd 230 ab, 230 ac, 230 ad, 234 bd, 234 cd 242 ab, 261 ad, 238 bc, 230 bd, 235 cd, 260 dd 236 ab, 231 ad, 249 bc, 231 bd, 243 dd 241 ab, 238 ac, 230 ad ae, 237 bc, 239 de, 230 bd 255 ab, 230 ac, 232 bc, 296 bb, 231 cc 238 ab, 235 ac, 236 ad, 238 bd, 238 cd 232 ab, 239 ac, 231 ad, 230 bd, 226 cd

167 b 166 a, 163 c 163 a, 179 d 149 a 152 a 181 a, 182 c, 182 d 206 b, 184 c, 181 d 162 c 160 a 177 a, 187 d 179 b, 174 c 178 b c 190 a, 180 b 192 b c d 163 c

a For the notation of the atoms in the clusters, see Figures 1-4. b The types of Ni atoms involved are indicated after the value; see Figures 1-4. The type of the Ni atom bound to the impurity X is indicated after the value; see Figures 2-4.

compounds concluded that this exchange-correlation potential furnishes very satisfactory accuracy, in particular when one considers the computational effort.40 For all open-shell systems, we checked the spin contamination of the underlying KohnSham determinant; the contamination of the ground states of all local minima of Ni5 in bare form and with an impurity was below 4%. The Kohn-Sham orbitals were represented by Gaussian-type basis sets contracted in generalized fashion: (15s11p6d) f [6s5p3d] for Ni,10,41 (9s5p2d) f [5s4p2d] for C and O, and (6s1p) f [3s1p] for H.42 The auxiliary basis set used in the LCGTO-FF-DF method to describe the Hartree part of the electron-electron interaction was derived from the orbital basis set in the usual fashion35 and augmented by five p-type and five d-type “polarization” exponents for each atom. These exponents were constructed as geometric series with a factor 2.5, starting with 0.1 and 0.2 for p- and d-exponents, respectively.35,43 Gas-phase Ni5 clusters with and without an impurity were optimized with the help of analytical energy gradients invoking various symmetry constraints (C4V, D3h, C3V, C2V, Cs or C1).44 The stability of bare clusters was characterized by the binding energy (BE) per Ni atom (in 3d94s1 configuration; Table 1):

BE ) -[Etot(Ni5) - 5E(Ni)]/5 For clusters with an impurity atom, we determined the binding energy of the impurity XdC, O, H with respect to the most stable isomer of bare Ni5:

BE ) -[Etot(XNi5) - Estab(Ni5) - E(X)] Thus, the larger a BE value, the more stable are the species Ni5 or XNi5, respectively. For the optimized configurations, we carried out a normal-mode analysis without symmetry constraints, to corroborate that the structures represent local minima of the potential energy surface (Table S1 of Supporting Information). For this purpose, the second-order energy derivatives were estimated as finite differences of analytic displacement gradients. In cases with at least one imaginary vibrational frequency, the original symmetry constraint was reduced and the structure was re-optimized. The vertical IP and EA values of bare and ligated clusters (Table 2) were estimated as differences of total energies (∆SCF

TABLE 2: Vertical Ionization Potential IP, Electron Affinity EA, HOMO Energy EHOMO and Spin Character of the HOMO (r - Majority, β - Minority), HOMO-LUMO Gap ∆EHL of Bare Ni5 and Ni5 Clusters with a H, C, or O Impurity Atom (Energies in eV) structuresa Ni5 exp. Ni5/C4V-sp Ni5/C3V-bp (µ-H)Ni5/Cs-sp1 (µ-H)Ni5/Cs-sp2 (µ3-H)Ni5/Cs-bp (η-H)Ni5/Cs-bp (η-H)Ni5/Cs-sp (µ4-C)Ni5/Cs-sp (µ3-C)Ni5/Cs-sp (η-C)Ni5/Cs-bp (η-C)Ni5/Cs-sp (µ3-O)Ni5/Cs-bp (µ-O)Ni5/Cs-bp (µ-O)Ni5/C1-sp (µ3-O)Ni5/Cs-sp (µ4-O)Ni5/Cs-sp (η-O)Ni5/Cs-sp

IP

EA

6.18 ( 0.24 6.22 ( 0.05d 6.17 ( 0.02e 6.64/6.46g 6.41/6.24g 6.95 6.56 6.45 6.77 7.08 6.59 6.69 6.73 7.07 6.53 6.80 6.90 6.54 6.34 7.04 c

1.57 ( 1.64 1.39 1.62 1.79 1.51 1.71 2.03 1.68 1.79 2.08 2.09 1.30 1.94 1.77 1.62 1.47 2.06

HOMO

∆HLb

-4.45 R -4.03 β -4.32 R -4.40 R -4.11 R -4.31 β -4.79 R -4.31 β -4.37 β -4.50 R -4.82 R -4.39 β -4.44 β -4.58 R -4.29 R -3.95 β -4.77 β

0.75 0.51 0.69 0.51 0.36 0.66 1.98 0.45 0.38 0.35 0.56 1.02 0.34 0.79 0.57 0.32 0.68

0.06f

a For the notation of the clusters with impurity atoms see Section 3.2; for the structures, see Figures 1-4. b HOMO-LUMO gap calculated within one spin manifold. c Ref 47. d Ref 48. e Ref 49. f Ref 53. g Calculated adiabatic IP.

procedure). We also determined atomic charges, both with a Mulliken analysis and by fitting the electrostatic potential45 (potential derived charges, PDC) (Tables S2-S5 of Supporting Information). We characterized the local spin s of an atomic center by the difference between majority and minority charge contributions of that center as obtained in a Mulliken analysis (Tables S2-S5 of Supporting Information). 3. Results and Discussion 3.1. Bare Cluster Ni5. We started our studies with the bare cluster Ni5 as reference, exploring four topologies: (i) a square pyramid (Ni5-sp, Figure 1a), (ii) a trigonal bipyramid (Ni5-bp, Figure 1b), (iii) a tetrahedral cluster with the fifth Ni atom bound to a Ni-Ni bond (Ni5-t); and (iv) a square-planar cluster with the fifth Ni atom bound to a Ni-Ni bond (Ni5-e). The

Ni5 Cluster with Single Impurity Atoms

Figure 1. Structures of the bare Ni5 cluster. Also shown are the labels for the different types of Ni atoms; see text and tables.

optimization of the square pyramid structure with C4V and C1 symmetry constraints lead to the same geometry, binding energy, and spin state of the cluster; the C4V symmetry of the ground state was confirmed by normal-mode analysis. The trigonal bipyramid structure was optimized with D3h, C3V, and Cs symmetry constraints. The bipyramidal structure Ni5/D3h-bp features one imaginary frequency in the vibrational analysis, while the structure with lower symmetry, Ni5/C3V-bp, represents an energy minimum.46 The last two isomers, Ni5-t and Ni5-e, were optimized in C2V and Cs symmetry constraints and are less stable than the square pyramid by 50 and 90 kJ/mol, respectively; however, each of these structures exhibit imaginary vibrational frequencies. Release of the symmetry constraints from the initial Ni5-e and Ni5-t geometries results in the square pyramidal and bipyramidal structures, respectively. Thus, the only stable structures of the neutral Ni pentamer are the clusters Ni5/C4V-sp and Ni5/C3V-bp. The stability of the obtained spin states of the bare Ni5 isomers was checked by single-point calculations (at fixed geometry) where the number of unpaired electrons, Ns, was varied by ( 2. In the square pyramid structure, the number of unpaired electrons in the ground state is Ns ) 6, while the trigonal bipyramid, Ni5/C3V-bp, has two unpaired electrons less, Ns ) 4. The square pyramid structure of Ni5 (Figure 1a) is 23 kJ/ mol more stable than the trigonal bipyramid structure (Figure 1b). As square pyramidal and triangular bipyramidal structures of Ni5 differ very little in energy, we reoptimized them using an uncontracted basis set (15s,11p,6d). We obtained the same structures; nearest-neighbor distances changed less than 1 pm. Also the relative stability of the two stable isomers remained the same as with the contracted basis set; the energy difference changed very little, 1 kJ/mol for the whole cluster. None of the modeled high-symmetry structures, Ni5/C4V-sp or Ni5/C3V-bp, has partially occupied degenerate spin orbitals. In both cases the HOMO is a doubly occupied e-type spinorbital. The Ni-Ni distances of the isomer Ni5/C4V-sp are 231 pm (Figure 1a and Table 1), close to the values in the tetrahedral structure of Ni4, 230 pm.32 In the bipyramidal structure, Ni5/ C3V-bp, the distances between the Ni atoms in the triangular base of the bipyramid are 229 pm, while the Ni-Ni distances involving one atom at an apex are longer, 236 and 237 pm (Figure 1b and Table 1). The tetrahedral Ni4/Cs cluster32 has a very small HOMOLUMO gap, 0.09 eV, but the gap of the most stable structure of Ni5, the cluster Ni5/C4V-sp, is much larger, 0.75 eV (Table 2). Ni 3d Mulliken populations dominate to more than 80% the six unpaired majority spin orbitals of the square pyramidal cluster Ni5/C4V-sp; the remaining contributions exhibit 4s character. The two highest occupied MOs have majority spin character and exhibit dominant 4s contributions of the four Ni atoms at the base of the pyramid, while the 4s AO of the apex

J. Phys. Chem. A, Vol. 111, No. 11, 2007 2069 atom Nia contributes to an unoccupied molecular orbital (LUMO+2). The apex Ni atom of Ni5/C4V-sp has the effective valence electronic configuration s0.72p0.34d8.93, the Ni atoms at the base of the pyramid have the configuration s0.93p0.24d8.84. A somewhat higher d population is determined for the Ni atoms of Ni5/C3Vbp which are s0.71p0.28d8.95 (base) and s1.04p0.12d8.94 (apex). In both structures of Ni5, the metal atoms with three nearest neighbors (apex atoms in bp and base atoms in sp) have higher 4s populations than the atoms with four neighbors (base atoms in bp and apex in sp). In both structures the spin population s per atom is higher on Ni atoms with three neighbors and smaller for atoms with four neighbors. The variation of s for different atoms is larger for the Ni5 square pyramid, 0.35 e, than that for the bipyramidal structure, 0.14 e (Table S2). The vertical ionization potential of the square pyramid structure of bare Ni5 cluster is 6.64 eV (Table 2), ∼0.4 eV higher than the experimental values 6.18 ( 0.24,47 6.22 ( 0.05,48 and 6.17 ( 0.02 eV.49 The vertical IP of Ni5/C3V-bp is smaller, 6.41 eV (Table 2), hence closer to the experimental values. As the calculated vertical IP value overestimates the experimental ionization potential, we calculated the adiabatic IP of the square pyramid structure without symmetry restrictions (Table 2). The resulting value, 6.46 eV, is 0.18 eV smaller than the vertical IP value because the structure of the final state cation differs from that of the initial state (square pyramid) which was used in the evaluation of the vertical IP. The same effect is calculated for Ni5/C3V-bp (Table 2). The structure of Ni5+ can be considered as intermediate between bipyramidal and square pyramid structures.50 The estimated adiabatic IP values of the square pyramid and bipyramidal structures are 6.46 and 6.24 eV, respectively. The former value is ∼0.3 eV higher than experiment, while the latter one essentially coincides with experiment.51 The vertical EA of the square pyramid structure, 1.64 eV (Table 2), is very close to the experimental value 1.57 ( 0.06 eV,53 while the vertical EA of the bipyramidal isomer Ni5/ C3V-bp, 1.39 eV, is ∼0.2 eV lower. Thus, both Ni5 isomers show reasonable agreement of calculated IP and EA values with experimental results and either of them could correspond to the experimentally studied Ni5 clusters. The calculated vertical IP and EA values of the stable Ni5 clusters are higher than the corresponding values calculated for the tetrahedral Ni4 cluster by 0.75 and 0.24 eV, respectively. This trend fits the experimental values of Ni5 which are also larger, by 0.47 (IP) and 0.07 eV (EA), than the corresponding experimental values of Ni4.49,53 The vibrational frequencies of the stable structures Ni5/C4Vsp and Ni5/C3V-bp fall into the intervals 78-312 cm-1 and 61334 cm-1, respectively (Table S1 of Supporting Information). The vibrational frequencies of the tetrahedron-like cluster Ni4 were calculated in the same region, 61-329 cm-1.32 Previous computational studies of bare Ni5 clusters yielded two isomers (Table 3), square pyramid and/or trigonal bipyramid structures with six and four unpaired electrons, respectively,21,54-56 as obtained in the present work (Table 1). Also, the relative stability of the two isomers had been calculated to be very similar; the overall energy difference was at most 20 kJ/mol. The local density approximation (LDA) combined with triple-ζ quality basis set54 gives a BE per Ni atom of 275 kJ/mol for the trigonal bipyramid and 271 kJ/mol for the square pyramid (Table 3). Calculations using the Perdew’s exchange57 and correlation39 functional yielded only 1 kJ/mol difference per atom between the two isomers21 with structures optimized at the highest symmetries, D3h and C4V, respectively. Recent DF

2070 J. Phys. Chem. A, Vol. 111, No. 11, 2007

Petkov et al.

TABLE 3: Results from Previous Calculations for the Cluster Ni5: Binding Energy per Ni Atom BE in kJ/mol, Number of Unpaired Electrons Ns, and Ni-Ni Distances in pm method, basis set LSDA, TZa P91/P86, TZb BLYP, DNPc

VWN, DZVP2e PW86/P86, DZVP2e GGA-NSe

symmetry, shape

BE

Ns

distances

D3h-bp C4V-sp D3h-bp C4V-sp D3h-bp C2V-bp C4V-sp C2V-sp Bp Bp Bp

275 271 213 214 341 343 346 347 296 241 197

4 6 4 6 4 4 6 6 4 4 4

231, 239 226, 228 231, 239 234, 235 d d 231, 236

a Ref 54. b Ref 21; P91/P86 denotes a combination of Perdew’s exchange, ref 57, and correlation, ref 39, functionals. c Ref 56, DNP denotes a double numerical basis with polarization functions where the 1s, 2s and 2p orbitals are kept “frozen.” d Averaged value. e Ref 55, PW86/P86 denotes a combination of the exchange functional suggested by Perdew and Wang, ref 59, and the correlation functional of ref 39.

modeling using the BLYP functional56 found the square pyramid isomer to be overall 20 kJ/mol more stable than trigonal bipyramid, where both structures had been optimized with a C2V symmetry constraint (Table 3). However, the calculated BE per Ni atom was rather high for a GGA exchange-correlation functional, 347 kJ/ mol,58 compared to calculations with other GGA functionals including the present results which yielded a BE per Ni atom of 231 kJ/mol. The Ni-Ni distances that we obtained for the bipyramidal structure are shorter than the values reported in previous works. The equatorial distances are up to 2 pm shorter; this holds for previous calculations with various exchange-correlation functionals (LDA,54 GGA21,39,55,57,59) while the bonds involving an apex atom differ by up to 3 pm (Table 3). All studies suggest close values for the Ni-Ni bonds at the base and the apex edges for the cluster with square pyramidal shape; the distances differ at most 2 pm. The average Ni-Ni distance, 227 pm, from LDA calculations54 is slightly shorter than that from our BP calculations, 231 pm, while optimization with P91/P8639,57 or BLYP38,60 functionals gives slightly longer distances,21,56 235 pm. Finally, we note that the reliability of the exchange-correlation functional and the basis set used in the present work for reproducing experimental values of binding energy and interatomic distances61 was recently confirmed for Ni2.40,62 As direct experimental information about the structure of small neutral nickel clusters is not available, we may compare our results with structural information derived from experiments with cations. Parks et al.63 used the saturation coverage of Nix cluster cations with N2 ligands to deduce the structure. For Ni5+ they found five and eight N2 molecules adsorbed at low and high N2 pressure, respectively, and concluded that the structure of the cluster should be a trigonal bipyramid. This agrees with our finding that the stable structure of Ni5+ resembles that of a distorted bipyramid with a single extended equatorial Ni-Ni distance (see endnote 50). However, the adsorption energy of a N2 molecule on a small Ni cluster, estimated at 116 kJ/mol per ligand for the neutral complex Ni3(N2)3 and at 86 kJ/mol per ligand for the corresponding cation,62 is much larger than the total energy difference between square pyramidal and bipyramidal conformations of the bare Ni5 cluster, 23 kJ/mol. Therefore, N2 coordination could easily change the structure of the metal moiety.

Figure 2. Stable structures of HNi5 clusters. Also shown are the labels for the different types of Ni atoms; see text and tables.

Figure 3. Stable structures of CNi5 clusters. Also shown are the labels for the different types of Ni atoms; see text and tables.

3.2. Clusters with Impurity Atoms. 3.2.1. Localization of Stable Minima. To locate the preferred positions of the impurity atoms H, C, and O on Ni5, we selected various initial positions of the heteroatoms at square pyramid and trigonal bipyramid metal frameworks (Figure S1 of the Supporting Information). To identify the bonding mode of the impurity, we use a nomenclature adapted from that common for inorganic complexes, that has served already well for XNi4 isomers.32 Direct coordination of an impurity atom to a single Ni atom is denoted by the prefix η (Figures 2d,e; 3c,d; 4f), coordination to a NiNi bond is denoted by the prefix µ (Figures 2a,b and 4b,c), and simultaneous coordination to three or four Ni atoms by the prefixes µ3 (Figures 2c, 3b, and 4a,d) or µ4 (Figures 3a and 4e), respectively. Thirteen initial structures of the clusters with impurities were constructed in the following way. For the square pyramid structure, the impurity atom X was located (i) in on-top fashion

Ni5 Cluster with Single Impurity Atoms at the apex and base Ni atoms, (η-X)Ni5-sp1 and (η-X)Ni5sp2, respectively; (ii) inside the cluster, (µ5-X)Ni5-sp; (iii) below the base of the pyramid, (µ4-X)Ni5-sp; (iv) at a side facet, (µ3X)Ni5-sp; (v) over a Ni-Ni bond at the base, (µ-X)Ni5-sp1; (vi) over a Ni-Ni bond at a side facet, (µ-X)Ni5-sp2 (see Supporting Information Figure S1a-g). For the bipyramidal structure, the impurity atom X was located (vii) on-top at the apex and base Ni atoms, (η-X)Ni5-bp1 and (η-X)Ni5-bp2, respectively; (viii) over a Ni-Ni bond involving an apex atom, (µ-X)Ni5-bp1; (ix) over a facet, (µ3-X)Ni5-bp1; and (x) over an equatorial Ni-Ni bond (µ-X)Ni5-bp2; and (xi) inside the cluster, (µ3-X)Ni5-bp2 (see Supporting Information Figure S1 h-m). All these structures were optimized applying pertinent symmetry constraints. To check the stability of the resulting structures, we carried out a normal-mode analysis without any symmetry restriction as described in Section 2 and proceeded at a lower symmetry constraint if necessary. The stability of the obtained spin states was checked by single-point calculations (at fixed geometry) where the number of unpaired electrons was varied by ( 2. For nearly degenerate structures, where we found a BE difference with respect to the initial multiplicity below 10 kJ/mol, we subsequently optimized the cluster geometry, keeping the spin multiplicity fixed. In the following, we discuss the geometry and some features of the electronic structure of the obtained stable minima. 3.2.2. Clusters with H Impurity. We identified five minima for an H impurity atom at the Ni5 cluster: (µ-H)Ni5/Cs-sp1, (µ-H)Ni5/Cs-sp2, (µ3-H)Ni5/Cs-bp, (η-H)Ni5/Cs-bp, and (ηH)Ni5/Cs-sp (Figure 2a-e). In the most stable structures, the first two of this series, the H atom bridges one of the two types of Ni-Ni bonds of the square pyramid, similarly to the coordination of an H atom at the tetrahedral Ni4 cluster.32 In (µ-H)Ni5/Cs-sp1 (Figure 2a) with BE(H) ) 273 kJ/mol, the impurity atom is coordinated at the base of the pyramid, while in the (µ-H)Ni5/Cs-sp2 isomer with BE(H) ) 266 kJ/mol, the H atom is bound at an edge involving an apex of the pyramid (Figure 2b). Both structures are sextets, i.e., the number of unpaired electrons decreases by one with respect to the bare square pyramidal Ni5 cluster. The binding energy of H in the most stable isomer, (µ-H)Ni5/Cs-sp1, is 12 kJ/mol smaller than the BE of this impurity atom at the cluster Ni4.32 Coordination of H to an edge at the base of the square pyramid shortens this Ni-Ni distance by 6 pm compared to the bare Ni5/C4V-sp cluster. A similar effect had been determined for an H impurity at tetrahedral Ni4.32 In the other isomer with bridge-coordinated H, (µ-H)Ni5/Cs-sp2, the impurity does not affect the length of the Nia-Nic bond to which it is coordinated (Figure 2b), but the Nia-Nib bond in trans position at the same apex is elongated by 11 pm. In addition to bridge-coordinated H, which was the only stable structure of the HNi4 cluster,32 for HNi5 we also found monoand three-coordinated positions to be local minima. In the isomer (µ3-H)Ni5/Cs-bp the heteroatom is three-coordinated. This structure has 5 unpaired electrons and a BE(H) value of 262 kJ/mol (Table 1, Figure 2c). The impurity H atom, coordinated at a facet of the trigonal bipyramid, is notably closer to the apex Nia, 163 pm, than to the two equatorial centers Nid, 179 pm. In the structures (η-H)Ni5/Cs-bp and (η-H)Ni5/Cs-sp (Figure 2d,e), the H atom is coordinated on top to an apex Nia atom of each of the two fundamental framework topologies, bipyramid and square pyramid, respectively. The corresponding BE(H) values are quite a bit smaller, 234 and 209 kJ/mol, respectively. The former cluster is stable with three unpaired electrons, by one less than bare Ni5/C3V-bp, while the latter isomer is an octet,

J. Phys. Chem. A, Vol. 111, No. 11, 2007 2071 i.e., it has one unpaired electron more than the corresponding bare Ni5/C4V-sp cluster. In four of the five reported structures, variation of the number of unpaired electrons by 2 resulted in a reduced stability of the isomer, by more than 25 kJ/mol. Only the cluster (µ3-H)Ni5/ Cs-bp has similar stability when optimized with three and five unpaired electrons: 259 and 262 kJ/mol, respectively (Table 1). This small energy difference does not allow us to assign the ground state configuration. However, on the basis of the estimated spin contamination, 18% for Ns ) 3 and 1% for Ns ) 5, we expect the higher spin multiplicity to represent the ground state. Extended Hu¨ckel calculations64 suggested also bridge coordination of H to be most stable for Ni5, yet at an edge involving the apex of trigonal bipyramidal Ni5. Spin-polarized DF-GGA studies, directed to hydrogen adsorption on different surfaces of metallic nickel65 suggested slightly smaller binding energies (by 2-18 kJ/mol) than calculated here for bridge coordinated H at Ni5. The BE(H) values were 252-264 kJ/mol for Ni(100) and 255-264 kJ/mol for Ni(111). Periodic slab model calculations also showed that H adsorption on 3-fold sites (fcc or hcp) of low-index single-crystal surfaces is 12-21 kJ/mol more stable than adsorption on bridge sites.65 This trend is opposite to the results reported here for H adsorption at Ni5, where the isomer with bridge-coordinated H is 11 kJ/mol more stable than the isomer with a 3-fold coordinated impurity. In the most stable structure (µ-H)Ni5/Cs-sp1 the HOMO is a majority spin orbital. Both HOMO energy, -4.32 eV, and the HOMO-LUMO gap, 0.69 eV, of this complex are very similar to the analogous quantities of bare Ni5/C4V-sp (Table 2). A notable deviation in effective electron configuration with respect to the bare cluster was found only for the 4s population of atoms Nib, bound to H, which decreases by 0.19 e. In all HNi5 structures, the H atom features a small spin, s ) -0.02 to 0.14 e. In the most stable structure (µ-H)Ni5/Cs-sp1, the maximum difference between the spins s on the various types of Ni atoms is reduced from 0.35 e in the bare cluster to 0.18 e in the ligated cluster (Tables S2, S3 of the Supporting Information). This reduction occurs due to smaller s values of the Nib and Nic atoms at the base of the square pyramid where H is bound in the ligated cluster, 1.00 and 1.07 e, while the corresponding s values in bare Ni5/Cs-sp are 1.27 e. The s of the apex Nia atoms, 0.89 e, remains almost at value in the corresponding bare cluster, 0.92 e. On the other hand, the second isomer (µ-H)Ni5/Cs-sp2 with H coordinated at the apex bond Nia-Nic, features a larger difference between s values of the Ni atoms in the cluster, 0.56 e, because the impurity is bound to the apex Nia atom and reduces that s value by 0.27 e. In the structure with a three-coordinated H atom (Ns ) 5) at the trigonal bipyramidal cluster Ni5/C3V-bp (Ns ) 4), the additional unpaired electron increases the atomic spin s mainly on the atoms Nib and Nic which are distant from the impurity. The calculated atomic charges for various conformations of the HNi5 cluster are shown in Table S3 of the Supporting Information. Both types of charges of the H atom of HNi5 were determined to be negative: PDC from -0.51 e to -0.39 e and Mulliken charges from -0.25 e to -0.18 e. Similarly to the trend calculated for the HNi4 cluster,32 the vertical ionization potential of the most stable structure with an H impurity, (µ-H)Ni5/Cs-sp1, is higher than for the bare nickel cluster, by 0.31 eV. In contrast, the IP of the second most stable structure, (µ-H)Ni5/Cs-sp2, is 0.08 lower than the vertical IP of the bare Ni5/C4V-sp cluster (Table 2). The calculated IP values (Figure 5a) roughly follow a trend to smaller values with

2072 J. Phys. Chem. A, Vol. 111, No. 11, 2007

Figure 4. Stable structures of ONi5 clusters. Also shown are the labels for the different types of Ni atoms; see text and tables.

Figure 5. Dependence of IP (a) and EA (b) on the coordination number of the impurity atom. Squares, HNi5; circles, CNi5; triangles, ONi5; straight line, bare Ni5.

increasing coordination number of H, but this correlation is not strict, as can be seen both for η and µ coordinated impurities. The calculated electron affinity of four of the HNi5 structures varies in the interval 1.51-1.79 eV, which is close to the value of the bare cluster, 1.64 eV (Table 2). The only larger change in the EA is calculated for the least stable complex, (η-H)Ni5/ Cs-sp, with EA ) 2.03 eV. The EA values also decrease

Petkov et al. qualitatively with the coordination number of the impurity (Figure 5b). As previously done for Ni4,32 we tried to rationalize how a hydrogen impurity affects the magnetism of a Ni5 moiety. We used a simple model which focuses on the HOMO and the LUMO of the bare metal cluster25-27 and assumes that the HOMO-LUMO gap of the cluster is small. In a simplified twoorbital three-electron picture, the 1s orbital of the H impurity interacts with the most appropriate filled orbital of the cluster, resulting in a pair of bonding and antibonding MOs of HNi5. If the energy of that antibonding MO is above that of the LUMO of the (bare) cluster, the extra electron of H will occupy the LUMO of the bare cluster. The spin multiplicity of the cluster will decrease (increase) if the LUMO of the bare cluster belongs to the minority (majority) spin manifold. If the antibonding orbital is situated below the HOMO of the bare metal cluster, then it will be filled by an electron of the HOMO of the (bare) cluster and the spin character of the HOMO will affect the number of unpaired electrons in HNi5. Both HOMO and LUMO orbitals of the reference bipyramidal structure Ni5/C3V-bp belong to the minority spin manifold. Thus, one expects a decrease of the number of unpaired electrons if the Ni5-H antibonding orbital lies above the LUMO, or an increase if the antibonding orbital lies below the HOMO. (ηH)Ni/Cs-bp follows the former situation (∆Ns ) -1), while (µ3-H)Ni5/Cs-bp, corresponds to the latter situation (∆Ns ) 1). Hence, the simple orbital model does not allow a clear prediction. HOMO and LUMO of the Ni5/C4V-sp isomer of the bare cluster belong to majority and minority spin manifolds, respectively. Hence, both variants of the simple orbital argument predict a decrease of the number of unpaired electrons. Indeed, we found both (µ-H)Ni5/Cs-sp1 and (µ-H)Ni5/Cs-sp2 isomers to be stable with Ns ) 5 whereas the bare Ni5/C4V-sp has Ns ) 6 (Table 1). However, the result for the structure (η-H)Ni5/Cssp (with Ns ) 7) does not fit to the above orbital argument. Thus, the latter case is another example that demonstrates the limitations of the simplified orbital model for rationalizing the spin multiplicity. In this context, note that compared to Ni4, which has a small HOMO-LUMO gap of 0.09 eV,32 Ni5 structures feature considerably larger gaps: 0.75 eV for Ni5/ C4V-sp and 0.51 eV for Ni5/C3V-bp (Table 2). 3.2.3. Clusters with C Impurity. We found four stable structures for the cluster CNi5, two of them with C at a terminal position, and one each with C three- and four-coordinated (Figure 3); the stability of these structures increases in that order. All structures have four unpaired electrons. In the most stable structure (µ4-C)Ni5/Cs-sp (Figure 3a), BE(C) is 660 kJ/mol, which is by 20 kJ/mol higher than the BE of C at the Ni4 cluster.32 Similarly to the 4-fold coordinated C atom at Ni4, the optimized structure of (µ4-C)Ni5/Cs-sp (Figure 3a) can be considered as insertion of the impurity in a Ni-Ni bond s either into an apex bond of the square pyramid or an equatorial bond of the trigonal bipyramid structure of Ni5. The bonds of the C impurity to the four neighboring Ni atoms, 181-182 pm (Table 1), are very similar to the distances determined for (µ4-C)Ni4, 181-184 pm. In fact, the cluster (µ4-C)Ni5/Cs-sp can formally be constructed from (µ4-C)Ni4 simply by adding a Ni atom to a facet of an open tetrahedron. In the complex (µ3-C)Ni5/Cs-sp, the BE is 646 kJ/mol. The impurity is bound at the base of the square pyramid, close to three of the Ni atoms, at 181-184 pm, while the distance to the fourth metal center, Nib, is 206 pm. In the last two local minima, the C atom is coordinated at the apex Nia of the square pyramid or to an equatorial Ni atom

Ni5 Cluster with Single Impurity Atoms of a bipyramid which has been distorted toward a square pyramid. In both cases the BE of C, 378 and 403 kJ/mol, is considerably lower than for the previous two isomers. As expected from the lower coordination of C, the C-Ni bond is rather short, 160-162 pm (Table 1). While the C impurity on Ni4 leads to a stabilization of the HOMO with respect to the bare cluster in all observed structures, in the case of CNi5 the stabilization of the HOMO depends on the coordination of the impurity atom. In the stable isomers with a high coordination number of C, (µ4-C)Ni5/Cs-sp and (µ3C)Ni5/Cs-sp, the HOMO energy is essentially the same as in the bare cluster Ni5/C4V-sp, while for η-coordination the stabilization depends on the shape of the Ni moiety. The main change in the effective electronic configuration of the Ni centers in the most stable isomer (µ4-C)Ni5/Cs-sp with respect to the bare cluster Ni5/C4V-sp is the reduction of the occupation of 4s AOs of the metal atoms bound to C by 0.06-0.14. The spin s in CNi5 isomers (Table S4 of the Supporting Information) is not equally distributed on the nonequivalent types of Ni atoms. In the structures (µ4-C)Ni5/Cs-sp and (µ3C)Ni5/Cs-sp the spin s on the C atom is low, -0.01 and 0.12 e, respectively, while in one of the structures with top-coordinated C atom (η-C)Ni5/Cs-sp the spin on C is higher (by absolute value), -0.45 e. In all CNi5 clusters Ni atoms bound to carbon show some spin quenching compared to the bare Ni5 cluster. In all locally stable structures of CNi5, the C center has a strong “carbidic” character, with PDCs ranging from -0.54 e for (η-C)Ni5/Cs-sp to -0.89 e and -0.90 e in the more stable structures (µ3-C)Ni5/Cs-sp and (µ4-C)Ni5/Cs-sp, respectively (Table S4 of the Supporting Information). In both latter complexes, the positive counter charge is almost equally distributed over the four Ni atoms bound to C, while the farther Ni atom is neutral or slightly negative. The presence of a C heteroatom in the cluster does not significantly change the IP of the two stable structures (µ4-C)Ni5/ Cs-sp and (µ3-C)Ni5/Cs-sp compared to the bare square pyramid cluster (Table 2). The EA values of these isomers of CNi5 are 0.04 and 0.15 eV higher, respectively, than the EA of the bare cluster, 1.64 eV. As for the H impurity complexes, the IP and EA values of the clusters with a C impurity tend to lower values with increasing coordination number of the impurity atom (Figure 5). However, comparing the complexes (η-C)Ni5/Csbp and (η-C)Ni5/Cs-sp, one can conclude that the structure of the Ni5 moiety and the coordination site of the impurity also affect the IP value. 3.2.4. Clusters with O Impurity. At variance with the ONi4 complex,32 where we determined only one stable structure, (µO)Ni4/C2V, for the Ni5 cluster we found six local minima with an O impurity. The coordination of O in these complexes varies between 1- and 4-fold. The most stable isomers are based on the trigonal bipyramid structure of the Ni5 moiety with the impurity coordinated to a facet, (µ3-O)Ni5/Cs-bp, or apex edge, (µ-O)Ni5/Cs-bp, (Figure 4a,b; Table 1). The former isomer is more stable with a BE(O) of 436 kJ/mol; for the isomer with bridge-coordinated O we calculated a smaller value, 412 kJ/ mol. Both isomers are quintets, as the bipyramidal isomer of bare Ni5. In the most stable structure, the Nia-O bond to the apex Ni atom is calculated at 177 pm, while the Nid-O bonds to the metal atoms at the base of the bipyramid are 10 pm longer. All Ni-Ni distances of the facet, where the impurity is coordinated, are extended by ∼24 pm compared to the bare cluster. The structural changes in the Ni5 moiety of (µ-O)Ni5/ Cs-bp with bridge-coordinated O are smaller; the Nib-Nic bond closest to O elongates only by 13 pm (Table 1, Figure 4b).

J. Phys. Chem. A, Vol. 111, No. 11, 2007 2073 In the other four isomers of the ONi5 complex the Ni5 moiety has a square pyramidal shape. The O impurity atom is coordinated at the base or a side of the pyramid, to a Ni-Ni bond of the base, or to a Ni atom of the base, forming the complexes (µ4-O)Ni5/Cs-sp (Figure 4e), (µ3-O)Ni5/Cs-sp (Figure 4d), (µ-O)Ni5/C1-sp (Figure 4c), or (η-O)Ni5/Cs-sp (Figure 4f), respectively. The BE of the impurity to the square pyramid increases with decreasing coordination number: 340, 390, and 402 kJ/mol for four-, three-, and two-coordinated O. However, the mono-coordinated O ligand of the cluster (η-O)Ni5/Cs-sp forms an exception with the lowest value, BE(O) ) 313 kJ/ mol. The complex (µ-O)Ni5/C1-sp has six unpaired electrons, as the bare Ni5/C4V-sp cluster, whereas all other complexes with an O impurity have 4 unpaired electrons. The Ni-O distances increase with increasing coordination number of O, from 163 pm for (η-O)Ni5/Cs-sp to 192 pm for (µ4-O)Ni5/Cs-sp (Table 1). The energy of the HOMO of four of the structures with an O impurity is similar to that of the most stable bare cluster; the difference ranges from -0.13 to 0.14 eV (Table 2). In the structure with highest coordination of the impurity, (µ4-O)Ni5/ Cs-sp, the energy of the HOMO is 0.50 eV higher than for the bare cluster, while in the isomer with on-top bonded O, (ηO)Ni5/Cs-sp (Figure 4f), the HOMO is stabilized by 0.32 eV. The effective configurations of the Ni atoms of the most stable isomer (µ3-O)Ni5/Cs-bp suggest a strong decrease of the 4s population of the centers Nia and Nid compared to the bare cluster, due to Ni-O bonding which results in a positive potential-derived charge on these centers (Table S5 of the Supporting Information, Mulliken charges are given in brackets), q(Nia) ) 0.23 [0.22] e and q(Nid) ) 0.30 [0.29] e, while the Nib and Nic atoms, far from O, are essentially neutral, q(Nib,c) ) -0.05 [0.02, -0.05] e. The high electronegativity of the O atom is reflected in its large negative charge, ranging from -0.71 [-0.62] e to -0.76 [-0.78] e on the clusters with µ and µ3 coordination of the ligand and from -0.52 [-0.78] e to -0.64 [-0.51] e for µ4 and η coordination, respectively. The O atom features a spin s of 0.01-0.25 e in the isomers with four unpaired electrons and 0.33 e in the structure (µ-O)Ni5/C1-sp with six unpaired electrons. That spin is associated with the O 2p contribution to the HOMO of the cluster. The remaining spin s is not equally distributed over the Ni atoms. Depending on the location of the O impurity on the cluster, the vertical IP of ONi5 isomers varies from 6.34 to 7.04 eV. The IP of the most stable structure (µ3-O)Ni5/Cs-bp is only 0.12 eV higher than the vertical IP of the bare Ni5/C3V-bp cluster. Calculated EA values vary from 1.30 to 2.06 eV. As can be seen in Figure 5, both the IP and EA values roughly decrease with the increasing coordination number of the impurity atom, similarly as for the clusters HNi5 and CNi5. The calculated binding energies of an oxygen impurity at a single-crystal nickel surface, calculated with respect to 1/2O2, are 163 kJ/mol for adsorption at a bridge site of Ni(111) and 221 kJ/mol for adsorption at an fcc site of that surface.66 For comparison with these values, we recalculated the BE of O in the relevant structures of the ONi5 cluster with respect to 1/2O2 using the calculated energy of the reaction 1/2O2(g) f O(g) of 196 kJ/mol; the resulting values were 216 kJ/mol for (µ-O)Ni5/ Cs-sp and 240 kJ/mol for (µ3-O)Ni5/Cs-bp. This comparison suggests that oxygen atoms bind stronger at small metal clusters than on a regular close-packed surface; this observation was already made for the Ni4 cluster.32 3.2.5. Detachment Energy of Ni Atoms from the Clusters. To evaluate the influence of the impurity atom on the stability of

2074 J. Phys. Chem. A, Vol. 111, No. 11, 2007

Petkov et al.

TABLE 4: Calculated and Experimentala Detachment Energy DE (kJ/mol) for a Ni Atom from the Ni5 Cluster with and without Impurity reaction

DE(calc)

DE(exp)

Ni5+ f Ni4+ + Ni Ni5 f Ni4 + Ni (µ-H)Ni5/Cs-sp1 f (µ-H)Ni4/C1(5) + Ni (µ4-C)Ni5/Cs-sp f (µ4-C)Ni4/C2V + Ni (µ3-O)Ni5/Cs-bp f (µ-O)Ni4/C2V + Ni

256 320 306 313 340

206 ( 23 256 ( 22

a

Ref 67.

small nickel clusters, we calculated the detachment energy (DE) of a Ni atom from the Ni5 cluster with impurity and compared it with the DE of bare Ni5 and Ni5+ clusters (Table 4). For both latter values, experimental estimates are available.67 For the evaluation of the calculated DE, we used the energies of the most stable structures of the corresponding clusters with five and four Ni atoms, reported here and in ref 32. The calculated DE of a Ni atom from neutral Ni5 is 320 kJ/mol, while a Ni atom is detached much easier from the cationic cluster Ni5+, where the DE is 256 kJ/mol. These values are about 50-64 kJ/mol overestimated68 compared to experimental results which were determined with an accuracy of (22 kJ/mol. However, the difference between calculated DE values for Ni5 and Ni5+, 64 kJ/mol, is quite close to the difference between the values derived from experiment, 50 kJ/mol. According to experiment67 the difference in the DE of the neutral and cationic clusters of the same nuclearity depends strongly on the size of the cluster. On the other hand, the presence of an impurity atom on the cluster modifies only slightly the detachment energy of a Ni atom, at most by 20 kJ/mol. H or O impurities facilitate the detachment of a Ni atom by 14 and 7 kJ/mol, respectively, compared to bare Ni5. A C impurity has the opposite effect: it slightly stabilizes the cluster as the DE value increases by 20 kJ/mol. On the basis of this result, one may speculate that a C impurity would foster the growth of a nickel cluster, while a H impurity would assist in its decomposition. Comparison of the DE of a Ni atom from the clusters with impurity with the BE of the impurity to Ni5 (Table 1) suggests that the cluster HNi5 will easier release the H impurity. The corresponding energies are 273 kJ/mol for the removal of H and 306 kJ/mol for the removal of a Ni atom. It requires less energy to remove a Ni atom from CNi5 and ONi5 clusters, with DE values of 340 and 313 kJ/mol, respectively, whereas the detachment of the impurity atom requires much more energy, 660 kJ/ mol for a C atom and 436 kJ/mol for a O atom. 3.2.6. EValuation of the ReactiVity of the Clusters. To evaluate the influence of impurity atoms on the reactivity of the Ni5 cluster, we estimated69 the total hardness70 η ≈ (IP - EA)/2 using the calculated values of IP and EA (Table 2). The hardness concept was previously applied to analyze the relative stability and the reactivity of small sodium and coinage metal clusters.71 The calculated values of the total hardness of both stable Ni5 structures and all Ni5X clusters are presented in Figure 6. The values are ordered according to the type of coordination of the impurity at the metal cluster, as in Figure 5. The two isomers of Ni5 yield essentially the same hardness, 2.5 eV. Impurity atoms hardly change the hardness (Figure 6) and hence the reactivity of the cluster; values vary within (10% around the result for Ni5. As both IP and EA tend to decrease with increasing coordination number of the impurity atom, the total hardness of the cluster changes neither with the coordination number nor the type of the impurity atom. For comparison we included in Figure 6 also the η values of the stable structures of Ni3, Ni4 and Ni6,32,62,72 calculated in

Figure 6. Estimated total hardness η of bare Nin clusters (n ) 3-6) and Ni5X clusters (X ) H, C, and O) with different coordination numbers of the impurity atom. Squares, HNi5; circles, CNi5; triangles, ONi5; rhombus, bare clusters.

analogous way. The hardness of Ni4 is 0.25 eV smaller than that of Ni5, while the values of Ni3 and Ni6 are larger, by 0.22 and 0.10 eV, respectively. Thus, the effect of impurities H, C, or O on the hardness of the cluster is comparable with the changes caused by addition to or removal of metal atoms from the cluster. The variations of the hardness of Nan, n ) 3-6, clusters (determined also from calculated IP and EA values) with their nuclearity were reported larger, 0.5-0.7 eV, and exhibited an odd-even oscillation, typical for sodium clusters.71a Following Pearson’s principle,70 Ni5 and Ni5X clusters should react preferably with reactants with similar total hardness, i.e., with η between 2 and 3 eV. These values are much smaller than the hardness of molecules considered as soft, as CO and N2 with η values above 7 eV. As the difference between the hardness of these ligand molecules and the cluster Ni5 is much larger than the variations of η from Ni5 to Ni5X, one should not expect substantial changes in ligand binding energy due to addition of a single impurity atom to Ni5. According to the estimated total hardness, the reactivity of the modeled metal clusters, both with and without impurities, is expected to be higher with softer reactants, e.g., with other metal clusters, suggesting a propensity to cluster aggregation. 4. Summary In agreement with earlier studies, the square pyramid and the trigonal bipyramid have been calculated as the more stable structures of Ni5. In the square pyramid structure, the bare Ni5 cluster has six unpaired electrons, Ns ) 6, while the trigonal bipyramid has two unpaired electrons less, Ns ) 4. In the square pyramid structure (Figure 1a), the cluster is 23 kJ/mol more stable than in the trigonal bipyramid structure (Figure 1b). The binding energy of the impurity (relative to an isolated impurity atom) increases in the order H (273 kJ/mol), O (436 kJ/mol), and C (660 kJ/mol; values quoted for the most stable isomers). Comparison with calculated atomization energies of H2 or O2 in the gas-phase suggests that the reactions 2 Ni5 + X2 f 2 Ni5X (X ) H, O) are exothermic with 90 kJ/mol (X ) H) and 480 kJ/mol (X ) O). The adsorption positions of H and C atoms at a nickel pentamer are essentially the same as those at a Ni4 tetramer. The preferred coordination of O on Ni5 is 3-fold, while on the smaller Ni4 cluster it was coordinated to a Ni-Ni bond. The structural distortions of the nickel moiety due to impurity coordination correlate with the BE of the impurity atom. The H impurity has the lowest BE and causes small changes in the

Ni5 Cluster with Single Impurity Atoms Ni-Ni bond lengths, at most (6 pm for the most stable isomer. The C atom, featuring a more than twice larger BE, induces a considerable distortion of the nickel cluster; one of the Ni-Ni bonds of Ni5 is broken in the energetically preferred structure. As expected, inclusion of an impurity atom H, C, or O in the cluster causes a partial oxidation of the metal atoms, as shown by a charge analysis, based either on potential derived or Mulliken charges. As a rough trend, the values of the IP and EA of the clusters with impurity atoms decrease with the coordination number of the impurity. In the most stable isomers of the clusters with C and O impurity the IP of the cluster is almost the same as for bare Ni5/C4V-sp, while the IP of the most stable HNi5 cluster is 0.3 eV higher. The EA values of the most stable isomers of the HNi5 and CNi5 clusters are close to the EA of the bare cluster, whereas the EA of the most stable ONi5 isomer is 0.3 eV lower than that of Ni5/C4V-sp. Comparison of the detachment energy of a Ni atom from the neutral and ionized Ni5 clusters suggested that it costs less energy to remove a Ni atom from the cationic species than from the neutral Ni5 cluster. Impurity atoms change the Ni removal energies in either way. A C impurity stabilizes the bonding of the Ni atoms in the cluster, whereas H or O impurities reduce the Ni detachment energy. From the calculated detachment energies one can conclude that it costs less energy to remove an H impurity from a HNi5 cluster than a Ni atom, while the clusters with a C or an O impurity lose much easier a Ni atom than the impurity atom. The estimates of the total hardness η as measure of the reactivity of the clusters with and without impurity atoms showed only small variations due to presence of an impurity atom ((10%) around the value of bare Ni5, η ) 2.50 eV. These variations are much smaller than the difference to the hardness of CO or N2 ligand molecules; therefore, one should not expect substantial differences in the binding energy of these ligands due to the presence of single impurity atoms on the nickel cluster. Acknowledgment. This work was supported by the National Science Fund and the University Research Fund (Bulgaria), Deutsche Forschungsgemeinschaft, and Fonds der Chemischen Industrie (Germany). Supporting Information Available: We provide a figure with initial structures of the clusters with impurities and tables with calculated vibrational frequencies, potential derived charges (PDC), Mulliken charges (MC), the corresponding spin densities per atom, and the electronic configurations of the isomers of bare Ni5 and the clusters with impurity atoms. This material is available free of charge via the Internet at http://pubs.acs.org. References and Notes (1) Armentrout, P. B. Annu. ReV. Phys. Chem. 2001, 52, 423. (2) Vayssilov, G. N.; Gates, B. C.; Ro¨sch, N. Angew. Chem., Int. Ed. 2003, 42, 1391. (3) Vayssilov, G. N.; Ro¨sch, N. Phys. Chem. Chem. Phys. 2005, 7, 4019. (4) Vajda, S.; Wolf, S.; Leisner, T.; Busolt, U.; Wo¨ste, L. H.; Wales, D. J. Chem. Phys. 1997, 107, 3492. (5) Goellner, J. F.; Neyman, K. M.; Mayer, M.; No¨rtemann, F.; Gates, B. C.; Ro¨sch, N. Langmuir 2000, 6, 2736. (6) Alexeev, O.; Gates, B. C. Top. Catal. 2000, 10, 273. (7) Sachtler, W. M. H.; Zhang, Z. AdV. Catal. 1993, 39, 129. (8) Kubelkova, L.; Vylita, J.; Brabec, L.; Drozdova, L.; Bolom, T.; Novakova, J.; Schulz-Ekloff, G.; Jaeger, N. I. J. Chem. Soc., Faraday Trans. 1996, 92, 2035. (9) Matveev, A. V.; Neyman, K. M.; Pacchioni, G.; Ro¨sch, N. Chem. Phys. Lett. 1999, 299, 603.

J. Phys. Chem. A, Vol. 111, No. 11, 2007 2075 (10) Ro¨sch, N.; Pacchioni, G. Inorg. Chem. 1990, 29, 2901. (11) Reuse, F. A.; Khanna, S. N. Chem. Phys. Lett. 1995, 234, 77. (12) Alonso, J. A. Chem. ReV. 2000, 100, 637. (13) Braunstein, P.; Oro, L. A.; Raithby, P. R. Metal Clusters in Chemistry; Wiley-VCH: Weinheim, 1999. (14) Pacchioni, G.; Kru¨ger, S.; Ro¨sch, N. in Ref. 13, pp 1392. (15) Ro¨sch, N.; Ackermann, L.; Pacchioni, G. Chem. Phys. Lett. 1992, 199, 275. (16) Ro¨sch, N.; Ackermann, L.; Pacchioni, G. Inorg. Chem. 1993, 32, 2963. (17) Ro¨sch, N.; Ackermann, L.; Pacchioni, G. J. Am. Chem. Soc. 1992, 114, 3549. (18) Pacchioni, G.; Chung, S.-C.; Kru¨ger, S.; Ro¨sch, N. Chem. Phys. 1994, 184, 125. (19) Apra, E.; Carter, E. A.; Fortunelli, A. Int. J. Quantum Chem. 2004, 100, 277. (20) Ruette, F.; Gonzalez, C. Chem. Phys. Lett. 2002, 359, 428. (21) Michelini, M. C.; Diez, R. P.; Jubert, A. H. Int. J. Quantum Chem. 2001, 85, 22. (22) Michelini, M. C.; Diez, R. P.; Jubert, A. H. Comp. Mat. Sci. 2004, 31, 292. (23) Reddy, B. V.; Nayak, S. K.; Khanna, S. N.; Rao, B. K.; Jena, P. J. Phys. Chem. A 1998, 102, 1748. (24) Kru¨ger, S.; Seemu¨ller, T. J.; Wo¨rndle, A.; Ro¨sch, N. Int. J. Quantum Chem. 2000, 80, 567. (25) Fournier, R.; Salahub, D. R. Surf. Sci. 1990, 238, 330. (26) Jones, N. O.; Beltran, M. R.; Khanna, S. N.; Baruah, T.; Pederson, M. R. Phys. ReV. B 2004, 70, 165406. (27) Ashman, C.; Khanna, S. N.; Pederson, M. R. Chem. Phys. Lett. 2003, 368, 257. (28) Yudanov, I. V.; Neyman, K. M.; Ro¨sch, N. Phys. Chem. Chem. Phys. 2004, 6, 116. (29) Neyman, K. M.; Inntam, C.; Gordienko, A. B.; Yudanov, I. V.; Ro¨sch, N. J. Chem. Phys. 2005, 122, 174705. (30) Bussai, C.; Kru¨ger, S.; Vayssilov, G. N.; Ro¨sch, N. Phys. Chem. Chem. Phys. 2005, 7, 2656. (31) Kru¨ger, S.; Bussai, C.; Genest, A.; Ro¨sch, N. Phys. Chem. Chem. Phys. 2006, 8, 3391. (32) Petkov, P. St.; Vayssilov, G. N.; Kru¨ger, S.; Ro¨sch, N. Phys. Chem. Chem. Phys. 2006, 8, 1282. (33) Ertl, G.; Kno¨zinger, H.; Weitkamp J. (Eds.) Handbook of Heterogeneous Catalysis Vol. 2, Wiley-VCH, Weinheim, 1997. (34) Hadjiivanov, K. I.; Vayssilov, G. N. AdV. Catal. 2002, 47, 307. (35) Dunlap, B. I.; Ro¨sch, N. AdV. Quantum Chem. 1990, 21, 317. (36) Belling, T.; Grauschopf, T.; Kru¨ger, S.; Mayer, M.; No¨rtemann, F.; Staufer, M.; Zenger, C.; Ro¨sch, N. High Performance Scientific and Engineering Computing, Lecture Notes in Computational Science and Engineering, Vol. 8, Bungartz, H.-J.; Durst, F.; Zenger, C., Ed.; Springer: Heidelberg, 1999; pp 439. (37) Belling, T.; Grauschopf, T.; Kru¨ger, S.; No¨rtemann, F.; Staufer, M.; Mayer, M.; Nasluzov, V. A.; Birkenheuer, U.; Hu, A.; Matveev, A. V.; Shor, A. M.; Fuchs-Rohr, M. S. K.; Neyman, K. M.; Ganyushin, D. I.; Kerdcharoen, T.; Woiterski, A.; Gordienko, A. B.; Majumder, S.; Ro¨sch, N. PARAGAUSS, version 3.0, Technische Universita¨t Mu¨nchen, 2004. (38) Becke, A. D. Phys. ReV. A 1988, 38, 3098. (39) Perdew, J. P. Phys. ReV. B 1986, 33, 8822; 1986, 34, 7406. (40) Furche, P.; Perdew, J. P. J. Chem. Phys. 2006, 124, 044103. (41) Wachters, A. J. H. IBM Tech Rep RJ584, 1969. (b) Wachters, A. J. H. Chem. Phys. 1970, 52, 1033. (42) (a) Van Duijneveldt, F. B. IBM Res. Report RJ945, 1971. The basis sets have been extended by two d polarization functions for C and O, 0.5591, 19.569, and 0.3074, 1.0757, respectively, and one p function for H, 1.0000. (b) Ferrari, A. M.; Neyman, K. M.; Ro¨sch, N. J. Phys. Chem. B 1997, 101, 9292. (43) Yudanov, I. V.; Sahnoun, R.; Neyman, K. M.; Ro¨sch, N. J. Chem. Phys. 2002, 117, 9887. (44) Matveev, A. V.; Mayer, M.; Ro¨sch, N. Comp. Phys. Comm. 2004, 160, 91. (45) Besler, B. H.; Merz, K. M.; Kollman, P. A. J. Comp. Chem. 1990, 11, 431. (46) The species Ni5/D3h-bp has the electronic configuration a2′1e′2e′′1 and thus will feature a Jahn-Teller distortion of first order. The degeneracy of the HOMO is removed when the symmetry constraints are reduced to C3V in the structure Ni5/C3V-bp, which yields an electronic configuration a21e2a11. (47) Parks, E. K.; Klots, T. D.; Riley, S. J. J. Chem. Phys. 1990, 92, 3813. (48) Watanabe, K. J. Chem. Phys. 1954, 22, 1564. (49) Knickelbein, M. B.; Yang, S.; Riley, S. J. Chem. Phys. 1990, 93, 94. (50) We started the optimization of Ni5+ from a square pyramid structure. During optimization the cluster distorted as the diagonals of the tetragonal base become different, 273 and 367 pm, which is to be compared to 327

2076 J. Phys. Chem. A, Vol. 111, No. 11, 2007 pm for the diagonals of the Ni5/C4V-sp structure. The resulting structure can also be considered as distorted trigonal bipyramid where one of the equatorial Ni-Ni bonds is considerably extended to 273 pm, while the other Ni-Ni distances are 233 pm (equatorial) and 231, 236 pm (axial). The deviation of the cationic cluster from both high symmetry structures can be rationalized as Jahn-Teller distortion. The HOMO, occupied by only one electron, originally is a degenerate e-type set, which splits upon distortion. The stability of the ionized species Ni5+ with respect to 4 Ni atoms and 1 Ni+ ion is 603 kJ/mol on average per mononuclear species. (51) The IP values were calculated also taking into account scalar relativistic effects as implemented in PARAGAUSS.52 The obtained values are: IP(vertical) ) 6.52 eV, IP(adiabatic) ) 6.29 eV for Ni5/C3V-bp; IP(vertical) ) 6.81 eV, IP(adiabatic) ) 6.63 eV for Ni5/C4V-sp. These relativistic values are by 0.05-0.17 eV higher than the non-relativistic results, and thus increase slightly the deviation from experiment compared to the non-relativistic approach. (52) Ro¨sch, N.; Kru¨ger, S.; Mayer, M.; Nasluzov, V. A. In Theoretical and Computational Chemistry, Vol. 4, Recent DeVelopments and Applications of Modern Density Functional Theory; Seminario, J. Ed.; Elsevier: Amsterdam, 1996; p 497. (53) Liu, S-R.; Zhai, H.-J.; Wang, L.-S. J. Chem. Phys. 2002, 117, 9758. (54) Michelini, M. C.; Diez, R. P.; Jubert, A. H. J. Mol. Struc. 1999, 490, 181. (55) Castro, M.; Jamorski, C.; Salahub, D. R. Chem. Phys. Lett. 1997, 271, 133. (56) Xie, Z.; Ma, Q.-M.; Liu, Y.; Li, Y.-C. Phys. Lett. 2005, A342, 459. (57) (a) Perdew, J. P. In Electronic Structure of Solids’91; Ziesche, P., Eschring, H., Eds.; Akademie: Berlin, 1991. (b) Perdew, J. P.; Chevary, J. A.; Vosko, S. H.; Jackson, K. A.; Pederson, M. R.; Singh, D. J.; Fiolhais, C. Phys. ReV. B 1992, 46, 6671. (58) This large values of BE, obtained with the BLYP functional, is specific to the work of Xie et al., ref 56. For example, the BE per Ni atom of Ni2, calculated with the functionals BLYP and BP, was reported to be 287 and 271 kJ/mol, respectively. These latter values are much higher than the corresponding experimental dissociation energy values of Ni2 per Ni atom, ref 61, 100 ( 1.0 kJ/mol, and the BE value calculated with the BP functional in the present work, 142 kJ/mol, ref 62. (59) Perdew, J. P.; Wang, Y. Phys. ReV. B 1986, 33, 8800. (60) Lee, C.; Yang, W.; Parr, R. G.; Phys. ReV. B 1988, 37, 785.

Petkov et al. (61) (a) Morse, M. O.; Hansen, G. P.; Langridge-Smith, P. R. R.; Zheng, L. S.; Geusic, M. E.; Michalopoulos, D. L.; Smalley. R. E. J. Chem. Phys. 1984, 80, 5400. (b) Pinegar, J. C.; Langenberg, J. D.; Arrington, C. A.; Spain, E. M.; Morse, M. D. J. Chem. Phys. 1995, 102, 666. (62) Aleksandrov, H. A.; Vayssilov, G. N.; Ro¨sch, N. J. Phys. Chem. A 2004, 108, 6127. (63) Parks, E. K.; Zhu, L.; Ho, J.; Riley, S. J. J. Chem. Phys. 1994, 100, 7206. (64) Curotto, E.; Matro, A.; Freeman, D. L.; Doll, J. D. J. Chem. Phys. 1998, 108, 729. (65) (a) Mattsson, T. R.; Wahnstrom, G.; Bengtsson, L.; Hammer, B. Phys. ReV. B 1997, 56, 2258. (b) Kresse, G.; Hafner, J. Surf. Sci. 2000, 459, 287. (c) Bhatia, B.; Sholl, D. S. J. Chem. Phys. 2005, 122, 204707. (66) Mittendorfer, F.; Eichler, A.; Hafner, J. Surf. Sci. 1999, 433-435, 756. (67) Lian, L.; Su, C. X.; Armentrout, P. B. J. Chem. Phys. 1992, 96, 7542. (68) The overestimation of the calculated detachment energy of a Ni atom is partially due to the fact that an Oh symmetry constraint causes an overestimation of the total energy of a Ni atom. Reduction of the symmetry results in a splitting of the d-shell and removes the degeneracy. The calculation of the atomic energy is complicated because in axial symmetry the 4s and the 3dz2 AO of Ni belong to the same irreproducible representation, a1, rendering it difficult to converge a configuration without holes below the HOMO. Thus, we estimated an upper limit of the atomic energy of Ni with 3d94s1 configuration in C4V and D2h symmetry (where the oneelectron spectra show holes). The corresponding total energies are 27 and 33 kJ/mol more stable than with Oh symmetry. Therefore, the overestimation of the experimental detachment energy for an Ni atom from the neutral Ni5 cluster is reduced from 64 kJ/mol to about 34 kJ/mol. (69) Parr, R. G.; Yang, W. Density Functional Theory of Atoms and Molecules; Oxford University Press: New York, 1989. (70) Pearson, R. G. J. Am. Chem. Soc. 1963, 85, 3533. (71) (a) Mineva, T.; Russo, N.; Toscano, M. Int. J. Quant. Chem. 2000, 80, 105. (b) Fernandez, E. M.; Soler, J. M.; Garzon, I. L.; Balbas, L. C. Int. J. Quant. Chem. 2005, 101, 740. (72) Petkov, P. St.; Vayssilov, G. N.; Kru¨ger, S.; Ro¨sch, N. 2007, in preparation.